Over the past years, automatic dialogue systems have received increasing attention. In addition to a speech recognizer, such systems include a natural language understanding (NLU) component. One of the most investigated approaches to NLU are rule-based methods as stochastic grammars which are often written manually. However, the sole usage of rule-based methods can turn out to be inflexible when extending or changing the application's domain. Therefore, techniques are desirable which help to reduce the manual effort when creating an NLU component for a new domain. In this paper we investigate an approach to NLU which is derived from the field of statistical machine translation. Starting from a bilingual annotated corpus, we describe the problem of NLU as the translation from a source to a target sentence. Experiments were performed on the TABA corpus which is a text corpus in the domain of a German train timetable information system.